Closed mandarjog closed 7 years ago
collected some data points for this PR -- in short, this seems solving the reported problem (istio/issues#100).
My setup is to simply deploy the istio-auth (without zipkin pod), bookinfo sample, and run wrk
to the productpage for 5 minutes.
The red line in the following graph shows the heap alloc bytes data (tracked by prometheus), with this PR (green is proxy -- unrelated, same in the following graphs):
Without this PR, typically heap allocation looks like
note that the spiky pattern happens after wrk run finishes -- still using memory for zipkin, we guess.
The number of goroutines,
Still have some increases since gRPC creates a goroutine for every request, but actually, the goroutine count looks like as follows without this PR
and note that, with the scale of the latter graph, the former goroutine "increase" looks like a flat line.
/lgtm
/lgtm
/approve
[APPROVALNOTIFIER] This PR is APPROVED
This pull-request has been approved by: douglas-reid, jmuk
The full list of commands accepted by this bot can be found here.
/test all [submit-queue is verifying that this PR is safe to merge]
Automatic merge from submit-queue.
Merging #1515 into master will not change coverage. The diff coverage is
n/a
.
@@ Coverage Diff @@
## master #1515 +/- ##
=======================================
Coverage 84.58% 84.58%
=======================================
Files 122 122
Lines 11999 11999
=======================================
Hits 10149 10149
Misses 1648 1648
Partials 202 202
Continue to review full report at Codecov.
Legend - Click here to learn more
Δ = absolute <relative> (impact)
,ø = not affected
,? = missing data
Powered by Codecov. Last update bbd5e27...6e2fca0. Read the comment docs.
This PR fixes goroutine proliferation issue with openzipkin-go. https://github.com/istio/issues/issues/100#issuecomment-339468537
After the PR is upstreamed, we will update the dependency.
https://github.com/mandarjog/zipkin-go-opentracing/pull/1
Release note:
This change is